Re: [Scilab-users] ode malfunction
Take care that the ode requires the function being at least continuous and differentiable If your input signal has singularities (points where it nos differentiable) you must stop the integration at this point and restart it just after taking the last state value as the intial point for the second run. Serge Steer Le 15/10/2013 17:46, Peter Hinow a écrit : Dear fellows, my goal is to solve numerically an ODE which contains a time-dependent parameter using 'ode'. I get a confusing result depending on whether the signal (it's called h) is on for 10 or 20 seconds. The two results are attached. Regardless of the model and the code, there should be a response in the variable w as soon as h switches from 0 to 1. I can provide any other information as needed. Should I use bugzilla to report this? I am not doing this very often, so I need some instruction how to proceed. Thank you and best regards, Peter ___ users mailing list users@lists.scilab.org http://lists.scilab.org/mailman/listinfo/users ___ users mailing list users@lists.scilab.org http://lists.scilab.org/mailman/listinfo/users
Re: [Scilab-users] ode malfunction
Serge, I suggest that an optional argument for ode would be a set of dates where the function is singular, so ode manages this by itself. What do you think? BR Adrien On 16/10/2013 03:46, Serge Steer wrote: Take care that the ode requires the function being at least continuous and differentiable If your input signal has singularities (points where it nos differentiable) you must stop the integration at this point and restart it just after taking the last state value as the intial point for the second run. Serge Steer Le 15/10/2013 17:46, Peter Hinow a écrit : Dear fellows, my goal is to solve numerically an ODE which contains a time-dependent parameter using 'ode'. I get a confusing result depending on whether the signal (it's called h) is on for 10 or 20 seconds. The two results are attached. Regardless of the model and the code, there should be a response in the variable w as soon as h switches from 0 to 1. I can provide any other information as needed. Should I use bugzilla to report this? I am not doing this very often, so I need some instruction how to proceed. Thank you and best regards, Peter ___ users mailing list users@lists.scilab.org http://lists.scilab.org/mailman/listinfo/users ___ users mailing list users@lists.scilab.org http://lists.scilab.org/mailman/listinfo/users -- Adrien Vogt-Schilb PhD Student (Cired) and Consultant (The World Bank) 1 202 473 79 80 ___ users mailing list users@lists.scilab.org http://lists.scilab.org/mailman/listinfo/users
Re: [Scilab-users] ode malfunction
Hello Peter, I think that your code might be useful. You might get some discontinuities when ode solver might get confused, or you have chosen the wrong solver. Regards, Radovan On 10/15/2013 5:46 PM, Peter Hinow wrote: Dear fellows, my goal is to solve numerically an ODE which contains a time-dependent parameter using 'ode'. I get a confusing result depending on whether the signal (it's called h) is on for 10 or 20 seconds. The two results are attached. Regardless of the model and the code, there should be a response in the variable w as soon as h switches from 0 to 1. I can provide any other information as needed. Should I use bugzilla to report this? I am not doing this very often, so I need some instruction how to proceed. Thank you and best regards, Peter ___ users mailing list users@lists.scilab.org http://lists.scilab.org/mailman/listinfo/users ___ users mailing list users@lists.scilab.org http://lists.scilab.org/mailman/listinfo/users